vicidial.org

One of our clients is having us set up Vicidial to increase the dialing capacity of their old dialer. We are doing this by using remote agents. When a call picks up it is connected to the remote agent through a full PRI cross connect onto the other dialer. We over write the value of the caller id used in Vicidial when we do this with the actual number that we dialed so that the other dialer will display the correct information for the agent. The agent is using the old dialers screens to status the calls. The problem that we are having is that some of the calls do not actually hangup. The system has 23 remote agents with a dial ratio of 7 to 1 and enough T1s that line capacity is not an issue. It has 2 32bit Xeon's with HT and 4 gigs of ram. We have watch the load on the system when it is dialing to make sure it does not get too high.

This is the same client that has the odd PRI lines that we have discussed here:
http://www.eflo.net/VICIDIALforum/viewt ... 0895#10895

We are not sure which side of the conversation actually did the hangup and if it is in fact the lines causing the problem or something else. One of the interesting things was that we checked through the DB for all of the numbers that were statused as XFER which was 36 and then grepped the asterisk log to see what happened on the calls. Four of those 36 calls never had DeadAGI called for them which seems odd, but not necessarily the problem because we had 9 out of 23 agents stuck not 4. We have just upgraded to asterisk-1.2.18 and zaptel-.2.17.1 maybe that will solve some of our issues. We will be running another test in the morning.

If this is on a Sangoma card, what wanpipe version are you using?

2.3.4-8

Oh well, I was going to mention that 2.3.4-7 had a problem with it that sounded a lot like your problem, but it was supposed to be fixed in -8

that still might have been part of the issue actually as we were using the 2.3.4-7 up till very recently.

We are going to be doing more testing on Monday.

So it looks like the problem is that in order to handle our funny lines we placed a 2 second wait in the hangup extension. This appears to cause some problems with the 1.2.18 version of asterisk (possibly 1.2.17 as well). They put "optimizations" into the hangup extension which appears to make it so that if the hangup extension takes too long not all of the code gets executed. Rather than wait in the hangup extension after the call we are waiting before the first agi call on the dial. Seems to fix the problems on low load. We will be doing large scale testing tomorrow.

Hi

is this problem still exits even we download from SVN from asterisk ?

ram

I would bet so because we have not had a chance to complain to the asterisk developers about this new behavior. I have not seen anything about this in a mailing list. However I have no real idea.